The Uncanny Valley Effect

Lay, Stephanie Claire (2015). The Uncanny Valley Effect. PhD thesis The Open University.



The Uncanny Valley Effect (UVE) first emerged as a warning against making industrial robots appear so highly human-like that they could unsettle the real humans around them. It proposed a specific pattern of negative emotional responses to entities that were almost but not quite human, and has been proposed as the reason why some entities such as dolls, mannequins and zombies may appear unsettling.

The aim of this thesis was to move beyond an anecdotal explanation to understand more about the perception of near-human faces, and how this compares to the perception of human and non-human faces. The aims were to explore the relationship between the human-likeness of faces and emotional responses to them, to understand reactions to and descriptions of near-human faces, to explore aspects of how near-human faces are processed and to explore whether mismatched emotional expressions might contribute to the perception of some near-human faces as eerie.

Five studies were carried out using face images whose human-likeness was systematically controlled or measured. A non-linear relationship between human-likeness and eeriness was found, but the near-human faces were not always the eeriest images. Near-human faces were found to be subject to the effects of inversion, and inversion was found to heighten perceptions of eeriness. Faces were created which contained mismatched emotional expressions, and the blends combining happy faces with angry or fearful eyes were rated as the most eerie. Incongruities between aspects of appearance or behaviour had been cited as explanations for the UVE in the past but this thesis presents the first evidence that differences in eeriness may result from incongruities between emotional expressions. Directions for future research have been suggested to explore these findings in a wider context and to understand more about the UVE.

Viewing alternatives

Download history


Public Attention

Altmetrics from Altmetric

Number of Citations

Citations from Dimensions

Item Actions